We’re about nine months into the Intel Xeon Scalable Processor (Skylake) server ramp and end customers have been making optimizations to their current software code and also rolling out new projects based on the new platform with its new capabilities. I got the chance to talk with Intel’s Navin Shenoy, EVP, and GM of the Data Center Group, on what looks to be an important vertical industry investment the company has made into healthcare. Shenoy says we are in the “golden age of data”, and with “30% of the world’s data in healthcare”, Intel can help. This makes perfect sense given Intel’s end to end in technologies from “process and analyze” (compute), “move” (networking) and “store” (storage) data. I wanted to share with you what I learned and relay some of the information from Intel’s SOLVE: Healthcare event held today in San Francisco.

Industry in crisis

As a former board chair for two of Austin’s premier hospitals, I am all too familiar with the healthcare industry’s challenges which have become human and societal challenges. At a $3T spend in the U.S., the healthcare industry is in crisis with rising costs, rising obesity, age and chronic health issues, and a focus on “break-fix” versus prevention. Insiders tell me that 75% of the $3T is spent on 2% of the population on “break-fix”, is getting worse every year, and that we’ll never catch up without some major changes. While the U.S. has its own exasperated set of issues, don’t be confused- this is a global issue.  While this may sound bleak, the good news is that some companies are leveraging the latest technologies to better address break-fix and predicting issues before they happen. While this doesn’t solve the prevention issue, it is a step in the right direction for treatment.

Intel’s role

Over the past 25 years, Intel has moved across the value chain from personal computer CPU maker to datacenter CPU to datacenter horizontal platform to horizontal solution with Select Solutions to now investing in vertical solutions. Intel has made big vertical investments in transportation, in financial verticals, and at today’s SOLVE: Healthcare event, Intel is ready to talk about its investments and early AI successes in healthcare. Intel stated in September 2017 that it has invested $1B in AI, and healthcare was identified as an important market for that investment. From what I saw today, the company and their partners are already showing results.

AI already happening in healthcare

It's easy to forget, but Artificial Intelligence (AI) is already happening today in healthcare and today, Intel was eager to outline what companies are doing with the new Xeon Scalable platform, particularly those using AI.  Here are some of the partners showcased at the SOLVE: Healthcare event.

Montefiore Health Systems has an AI platform called PALM (Patient-centered Analytics and Learning Machine) that ICUs are using to predict when patients might take a turn for the worse so it might be preemptively avoided.  The goal is to shorten ICU stays which the company said costs from $13K to $43K per average stay of two weeks. Its ultimate goal is to predict chronic conditions two years before they occur adding socioeconomic and genetics data.  There’s still the issue remaining of “now what”, but prevention platforms such as Wellsmith, I believe, are successfully starting to tackle this with healthcare providers.

Stanford University is involved in a number of different efforts to utilize AI and machine learning to improve and quicken the process of medical imaging for tumors. This makes total sense as machine learning is excellent at image detection, recognition, and classification. Stanford says it is doing this with Xeon cluster and have shortened image classification from 45 minutes to one minute. Stanford talked about how much safer this made the process for infant patients who needed to be sedated or intubated during imaging.

ICON plc wants to reduce the dependency on clinic visits and paper records for pharmaceutical trials, by instead gathering patient data from sensors and wearables. This seems very straightforward, but we aren’t talking about FitBits here- this requires medical-grade accuracy, compliance, and oversight. These hurdles are why there is such a separation today between health and medical industries. The Michael J. Fox Foundation and Teva Pharmaceutical said they are using the Intel Pharma Analytics Platform across dozens of clinical trials with 1,000s of patients.

Also highlighting their Xeon Scalable-based healthcare technologies were:

• AccuHealth for remote patient monitoring

• Broad Institute for genome analysis

• Diaceutics using AI to improve diagnostics

• Dr. Hazel using AI to detect skin cancer

• GE Healthcare using AI to improve medical imaging

• Harvard Medical School using AI to predict tumor response to chemotherapy

• Intermountain Healthcare using AI to analyze DNA to provide custom cancer treatments

• Mayo Clinic using AI to sift through big data to provide personalized care

• OptumLabs using AI to predict Alzheimer’s and dementia

• Princeton Neuroscience Institute using AI to provide brain stimulus to modify brain states

• UCSF using DL (deep learning) to predict osteoarthritis onset and intervene sooner

All of these new solutions (which, by the way, are the result of partnerships with Intel) are big leaps forward. However, the surface has only barely been scratched in terms of what’s possible. Most of these solutions focus on break-fix or break-fix prediction which is important, but I’d like to see prevention as the next focus of technology.

What Intel is bringing to the table

Within AI, the branch of deep learning, in particular, holds a lot of promise for healthcare. Deep learning, essentially, is the use of neural networks to train and infer insights from massive pools of complex and unstructured data. GPUs and rightly NVIDIA has gotten all the glory lately when it comes to AI, deep and machine learning. This ignores the fact that other companies have been doing AI for a decade on server processors, though not necessarily ML or DL. Are you familiar with the Amazon.com recommendation engine? You got it, driven by a server CPU as outlined at the Xeon Scalable launch according to an Amazon.com. There are healthcare deep learning applications that Intel says perform better on CPUs than GPUs.

Intel says GPUs are currently limited in their ability to perform healthcare function in part because of memory—the chips themselves can only hold 12 to 16 GB, which simply isn’t enough to run some of the massive AI models healthcare applications, such as imaging, often require. Additionally, while an organization would need an accelerator if it were training AI models 24/7, typically all that is needed is the ability to train and deploy individual models. This train versus deploy is an interesting argument and one that Amazon.com makes sense for certain situations.

With all this in mind, Intel is positioning its Xeon Scalable processor as the “ideal computational foundation” for healthcare organizations performing AI workloads—citing its ability to scale up quickly and flex in accordance to fluctuating business needs. Xeon Scalable processors can provide a terabyte of onboard memory, which can provide a higher level of resolution for medical imaging purposes than what can be achieved with GPUs. Additionally, Intel says the flexibility of the Xeon Scalable family allows for the infrastructure to dedicate more resources to other workloads in addition to those that require AI. This is a fair argument and Intel never says that today GPUs aren’t involved anywhere in the workflow.

Another product Intel will be bringing to the deep learning table in the future is its Nervana Neural Network Processor (NNP). NNP is focused on training and Intel says that it will allow healthcare organizations to increase their capacity to process and analyze large amounts of data. The company says this will have a myriad of positive impacts for healthcare, including but not limited to the fields of medical imaging, research, and early diagnosis of diseases. NNP is sampling now and 2018 is the year for software development with broader deployments in 2019. This is NVIDIA’s bread and butter in the datacenter, so watch this space.

Wrapping up

Intel is one of the few companies that can develop end-to-end healthcare technologies from the datacenter to the cloud, to the edge, and across compute, storage and the network. With this healthcare AI focus area, it looks to me that Intel is not only doing a lot of societal “good” but advancing up the value chain as well.

Intel is challenging a common worldview by demonstrating CPUs as an important AI technology. According to Intel’s healthcare partners, this is real and they are using Xeon to perform deep and machine learning workloads. Intel never says GPUs aren’t helping in healthcare- they are, but Xeon Scalable and Intel, based on what I saw today, are providing a lot of value and leadership right now.

Strategically, it’s smart for Intel to move up the value chain to vertical solutions like healthcare, financial services, and transportation as it’s harder to compete with and typically provides higher margins. In addition, by getting a vertical industry’s attention and investing heavily in it, as long as Intel shows up with the goods, they have a higher chance of making the sale.  I like this move.

Note: Moor Insights & Strategy writers and editors may have contributed to this article.

Disclosure: Moor Insights & Strategy, like all research and analyst firms, provides or has provided paid research, analysis, advising, or consulting to many high-tech companies in the industry, including Advanced Micro Devices, Apstra, ARM Holdings, Bitfusion, Cisco Systems, Dell EMC, Diablo Technologies, Echelon, Ericcson, Frame, Gen Z Consortium, Glue Networks, GlobalFoundries, Google (Nest), HP Inc. Hewlett Packard  Enterprise, Huawei Technologies, IBM, Jabil Circuit, Intel, Interdigital, Konica Minolta, Lenovo, Linux Foundation, MACOM (Applied Micro), MapBox, Mavenir, Mesosphere, Microsoft, National Instruments, NOKIA (Alcatel Lucent), Nortek, NVIDIA, ONUG, OpenStack Foundation, Peraso, Portworx, Protequus, Pure Storage, Qualcomm, Rackspace, Rambus, Red Hat, Samsung Technologies, Silver Peak, SONY, Springpath, Sprint, Stratus Technologies, TensTorrent, Tobii Technology, Synaptics, Verizon Communications, Vidyo, Xilinx, Zebra, which may be cited in this article.